Добавил:
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:
Professional C++ [eng].pdf
Скачиваний:
284
Добавлен:
16.08.2013
Размер:
11.09 Mб
Скачать

Chapter 21

//

//Remove students who are on the dropped list.

//Iterate through the dropped list, calling remove on the

//master list for each student in the dropped list.

//

for (list<string>::const_iterator it = droppedStudents.begin(); it != droppedStudents.end(); ++it) { allStudents.remove(*it);

}

// Done!

return (allStudents);

}

Container Adapters

In addition to the three standard sequential containers, the STL provides three container adapters: the queue, priority_queue, and stack. Each of these adapters is a wrapper around one of the sequential containers. The intent is to simplify the interface and to provide only those features that are appropriate for the stack or queue abstraction. For example, the adapters don’t provide iterators or the capability to insert or erase multiple elements simultaneously.

The container adapters’ interfaces may be too limiting for your needs. If so, you can use the sequential containers directly or write your own, more full-featured, adapters. See Chapter 26 for details on the adapter design pattern.

queue

The queue container adapter, defined in the header file <queue>, provides standard “first-in, first-out” (FIFO) semantics. As usual, it’s written as a class template, which looks like this:

template <typename T, typename Container = deque<T> > class queue;

The T template parameter specifies the type that you intend to store in the queue. The second template parameter allows you to stipulate the underlying container that the queue adapts. However, the queue requires the sequential container to support both push_back() and pop_front(), so you only have two built-in choices: deque and list. For most purposes, you can just stick with the default deque.

Queue Operations

The queue interface is extremely simple: there are only six methods plus the constructor and the normal comparison operators. The push() method adds a new element to the tail of the queue, and pop() removes the element at the head of the queue. You can retrieve references to, without removing, the first and last elements with front() and back(), respectively. As usual, when called on const objects, front() and back() return const references, and when called on non-const objects they return non-const (read/write) references.

588

Delving into the STL: Containers and Iterators

pop() does not return the element popped. If you want to retain a copy, you must first retrieve it with front().

The queue also supports size() and empty(). See the Standard Library Reference resource on the Web site for details.

Queue Example: A Network Packet Buffer

When two computers communicate over a network, they send information to each other divided up into discrete chunks called packets. The networking layer of the computer’s operating system must pick up the packets and store them as they arrive. However, the computer might not have enough bandwidth to process all of them at once. Thus, the networking layer usually buffers, or stores, the packets until the higher layers have a chance to attend to them. The packets should be processed in the order they arrive, so this problem is perfect for a queue structure. Following is a small PacketBuffer class that stores incoming packets in a queue until they are processed. It’s a template so that different layers of the networking layer can use it for different kinds of packets, such as IP packets or TCP packets. It allows the client to specify a max size because operating systems usually limit the number of packets that can be stored, so as not to use too much memory. When the buffer is full, subsequently arriving packets are ignored

#include <queue> #include <stdexcept> using std::queue;

template <typename T> class PacketBuffer

{

public:

//

//If maxSize is nonpositive, the size is unlimited.

//Otherwise only maxSize packets are allowed in

//the buffer at any one time.

//

PacketBuffer(int maxSize = -1);

//

//Stores the packet in the buffer.

//Throws overflow_error is the buffer is full.

void bufferPacket(const T& packet);

//

//Returns the next packet. Throws out_of_range

//if the buffer is empty.

//

T getNextPacket() throw (std::out_of_range);

protected:

queue<T> mPackets; int mMaxSize;

private:

// Prevent assignment and pass-by-value.

589

Chapter 21

PacketBuffer(const PacketBuffer& src); PacketBuffer& operator=(const PacketBuffer& rhs);

};

template <typename T> PacketBuffer<T>::PacketBuffer(int maxSize)

{

mMaxSize = maxSize;

}

template <typename T>

void PacketBuffer<T>::bufferPacket(const T& packet)

{

if (mMaxSize > 0 && mPackets.size() == static_cast<size_t>(mMaxSize)) {

// No more space. Just drop the packet. return;

}

mPackets.push(packet);

}

template <typename T>

T PacketBuffer<T>::getNextPacket() throw (std::out_of_range)

{

if (mPackets.empty()) {

throw (std::out_of_range(“Buffer is empty”));

}

//Retrieve the head element. T temp = mPackets.front();

//Pop the head element. mPackets.pop();

//Return the head element. return (temp);

}

A practical application of this class would require multiple threads. However, here is a quick unit testlike example of its use:

#include “PacketBuffer.h” #include <iostream> using namespace std;

class IPPacket {};

int main(int argc, char** argv)

{

PacketBuffer<IPPacket> ipPackets(3);

ipPackets.bufferPacket(IPPacket());

ipPackets.bufferPacket(IPPacket());

ipPackets.bufferPacket(IPPacket());

ipPackets.bufferPacket(IPPacket());

590

Delving into the STL: Containers and Iterators

while (true) { try {

IPPacket packet = ipPackets.getNextPacket(); } catch (out_of_range&) {

cout << “Processed all packets!” << endl;

break;

}

}

return (0);

}

priority_queue

A priority queue is a queue that keeps its elements in sorted order. Instead of a strict FIFO ordering, the element at the head of queue at any given time is the one with the highest priority. This element could be the oldest on the queue or the most recent. If two elements have equal priority, their relative order in the queue is FIFO.

The STL priority_queue container adapter is also defined in <queue>. Its template definition looks something like this (slightly simplified) one:

template <typename T, typename Container = vector<T>, typename Compare =

less<T> >;

It’s not as complicated as it looks! You’ve seen the first two parameters before: T is the element type stored in the priority_queue and Container is the underlying container on which the priority_queue is adapted. The priority_queue uses vector as the default, but deque works as well. list does not work because the priority_queue requires random access to its elements for sorting them. The third parameter, Compare, is trickier. As you’ll learn more about in Chapter 22, less is a class template that supports comparison of two objects of type T with operator<. What this means for you is that the priority of elements in the queue is determined according to operator<. You can customize the comparison used, but that’s a topic for Chapter 22. For now, just make sure that you define operator< appropriately for the types stored in the priority_queue.

The head element of the priority queue is the one with the “highest” priority, by default determined according to operator< such that elements that are “less” than other elements have lower priority.

Priority Queue Operations

The priority_queue provides even fewer operations than does the queue. push() and pop() allow you to insert and remove elements respectively, and top() returns a const reference to the head element.

top() returns a const reference even when called on a non-const object. The priority_queue provides no mechanism to obtain the tail element.

pop() does not return the element popped. If you want to retain a copy, you must first retrieve it with top().

591

Chapter 21

Like the queue, the priority_queue supports size() and empty(). However, it does not provide any comparison operators. The Standard Library Reference resource on the Web site for details.

This interface is obviously limited. In particular, the priority_queue provides no iterator support, and it is impossible to merge two priority_queues.

Priority Queue Example: An Error Correlator

Single failures on a system can often cause multiple errors to be generated from different components. A good error-handling system uses error correlation to avoid processing duplicate errors and to process the most important errors first. You can use a priority_queue to write a very simple error correlator. This class simply sorts events according to their priority, so that the highest-priority errors are always processed first. Here is the class definition:

#include <ostream> #include <string> #include <queue> #include <stdexcept>

// Sample Error class with just a priority and a string error description class Error

{

public:

Error(int priority, std::string errMsg) : mPriority(priority), mError(errMsg) {}

int getPriority() const {return mPriority; } std::string getErrorString() const {return mError; }

friend bool operator<(const Error& lhs, const Error& rhs);

friend std::ostream& operator<<(std::ostream& str, const Error& err);

protected:

int mPriority; std::string mError;

};

// Simple ErrorCorrelator class that returns highest priority errors first class ErrorCorrelator

{

public: ErrorCorrelator() {}

//

// Add an error to be correlated.

//

void addError(const Error& error);

//

// Retrieve the next error to be processed.

//

Error getError() throw (std::out_of_range);

protected:

std::priority_queue<Error> mErrors;

592

Delving into the STL: Containers and Iterators

private:

// Prevent assignment and pass-by-reference. ErrorCorrelator(const ErrorCorrelator& src); ErrorCorrelator& operator=(const ErrorCorrelator& rhs);

};

Here are the definitions of the functions and methods.

#include “ErrorCorrelator.h” using namespace std;

bool operator<(const Error& lhs, const Error& rhs)

{

return (lhs.mPriority < rhs.mPriority);

}

ostream& operator<<(ostream& str, const Error& err)

{

str << err.mError << “ (priority “ << err.mPriority << “)”; return (str);

}

void ErrorCorrelator::addError(const Error& error)

{

mErrors.push(error);

}

Error ErrorCorrelator::getError() throw (out_of_range)

{

//

//If there are no more errors, throw an exception.

if (mErrors.empty()) {

throw (out_of_range(“No elements!”));

}

//Save the top element.

Error top = mErrors.top();

//Remove the top element. mErrors.pop();

//Return the saved element. return (top);

}

Here is a simple unit test showing how to use the ErrorCorrelator. Realistic use would require multiple threads so that one thread adds errors, while another processes them.

#include “ErrorCorrelator.h” #include <iostream>

using namespace std;

int main(int argc, char** argv)

{

ErrorCorrelator ec;

593